摘要
在基于双能X射线透射技术的废金属分选中,识别效果很大程度上受所选物料特征影响。已有废金属物料识别算法所用特征不够全面且各特征数据间冗余性较大,识别准确度不高。针对该问题,充分考虑特征间冗余性和特征与类别间关联性提出相关性特征选择方法(CFS),在众多物料特征中确定由特征I_(H)、I_(L)、x、T_(ML)和T_(MH)组成的最优特征子集。通过采用不同物料特征组合方法进行对比实验,并选择最优特征子集下K-最近邻(KNN)作为最优分类模型。实验结果表明:基于相关性特征选择K-最近邻(CFS-KNN)的废有色金属物料识别分类方法与已有分类方法相比用相对较少的特征获得更高的识别准确度(96.13%)。
In the separation of scrap metal based on dual-energy X-ray transmission technology,the recognition effect is largely affected by the features of the selected materials.The features used in the existing scrap metal material recognition algorithms are not comprehensive enough,and the redundancy among feature data is large,so the recognition accuracy is not high.To solve this problem,this paper gives full consideration to the redundancy between features and the correlation between features and categories,and puts forward a correlation-based feature selection method(CFS)to determine the optimal feature subset consisting of features I_(H)、I_(L)、x、T_(ML) and T_(MH) among many material features.Comparative experiments were carried out by using different material feature combination methods,and K nearest neighbor(KNN)under the optimal feature subset was selected as the optimal classification model.The experimental results show that:compared with the existing classification methods,the identification and classification method of waste nonferrous metals materials based on the correlation-based feature selection and K nearest neighbor(CFS-KNN)has higher identification accuracy(96.13%)with relatively few features.
作者
陈煜昊
叶文华
徐祥
符杰
CHEN Yuhao;YE Wenhua;XU Xiang;FU Jie(School of Mechatronic Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China;Jiangsu Huahong Technology Co.,Ltd.,Wuxi 214400,China)
出处
《有色金属工程》
CAS
北大核心
2023年第1期86-92,共7页
Nonferrous Metals Engineering
基金
江苏省重点研发计划项目(BE2020779)。
关键词
特征选择
相关性
K-最近邻
废金属
分类
feature selection
correlation
K Nearest neighbor
scrap metal
classification
